I put NVIDIA’s new RTX 5090 through real-world testing to see what the “world’s fastest gaming GPU” actually feels like, and the answer is: it’s an absolute weapon… with some pretty ugly tradeoffs. On paper it’s a technological marvel—92 billion transistors, 32GB of VRAM, and a shockingly tiny PCB for a flagship card. In practice, I compared it directly against the RTX 4090 across a stack of games, and the uplift is much more mellow than last gen: roughly ~30% on average, with standouts like Red Dead Redemption 2 (~41%) and God of War (~38%). The bigger story is power. Where the 4090 often didn’t pull that much more than a 3090 despite being ~74% faster, the 5090 is regularly sitting around 480–500W in games and can slam into its 575W TDP. That’s a colossal jump, and it really doesn’t look like Blackwell is delivering meaningful efficiency gains here. The one genuinely new “experience” is Multi Frame Generation (2x/3x/4x), which can make path-traced Cyberpunk at 4K hit 240fps—but you’re trading for motion artifacts, and I still wouldn’t recommend it for anything even mildly competitive due to latency when normalized to the same frame rate. The cooler, though, is impressively competent for how much power this thing chugs.